test(agent-sdk): use GAIA_TEST_MODEL env var; set Llama in CI

The Agent SDK integration test hardcoded DEFAULT_MODEL_NAME (now
Gemma-4-E4B-it-GGUF) but the GitHub-hosted Windows CI runner only
pulls Llama-3.2-3B-Instruct-Hybrid, causing HTTP 422 from the server.

- tests/test_agent_sdk.py: read model from GAIA_TEST_MODEL env var,
  falling back to DEFAULT_MODEL_NAME so local runs still work
- test_agent_sdk.yml: set GAIA_TEST_MODEL=Llama-3.2-3B-Instruct-Hybrid
  before running the test suite to match the pulled model

Author: itomek

.github/workflows/test_agent_sdk.yml

@@ -145,6 +145,8 @@ jobs:
 
           REM Run the comprehensive integration test suite
           set PYTHONIOENCODING=utf-8
+          REM Use the model that was pulled above (overrides DEFAULT_MODEL_NAME=Gemma-4-E4B)
+          set GAIA_TEST_MODEL=Llama-3.2-3B-Instruct-Hybrid
           python tests\test_agent_sdk.py
           set integration_exit=%ERRORLEVEL%
 

tests/test_agent_sdk.py

@@ -9,6 +9,7 @@
 These tests require a running Lemonade server and test actual LLM interactions.
 """
 
+import os
 import sys
 import time
 import unittest
@@ -39,7 +40,7 @@ def setUpClass(cls):
         print(f"{'='*60}")
 
         cls.server_url = "http://localhost:13305"
-        cls.model = DEFAULT_MODEL_NAME
+        cls.model = os.environ.get("GAIA_TEST_MODEL", DEFAULT_MODEL_NAME)
         cls.timeout = 30  # seconds
 
         # Verify server is running